Context-Aware Gujarati Cricket Text Processing with LSTM-Based Word Generation

Abstract

This research explores the development of a Gujarati text generator specializing in cricket-related conversations. With the increasing need for domain-specific NLP applications in regional languages, this study aims to bridge the gap in automated text generation for Gujarati sports content. Utilizing a Long Short-Term Memory (LSTM)- based deep learning model, the system is trained on a limited dataset of Gujarati cricket news and commentary. The model is fine-tuned with beam search and temperature scaling to improve coherence and contextual relevance. The results demonstrate promising accuracy, with potential applications in sports journalism, chatbots, and content automation. Future work aims to integrate attention mechanisms and transformers-based architecture to enhance contextual understanding and output fluency.